the impact of intellectual heterogeneity on academic ... · a quantile regression approach is...
Post on 06-Apr-2018
218 Views
Preview:
TRANSCRIPT
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 1
The impact of intellectual heterogeneity
on academic performance in business education
Agnieszka Bielinska-Kwapisz
Montana State University, Bozeman
F. William Brown
Montana State University, Bozeman
ABSTRACT
This study extends previous lines of research which have identified academic
achievement determinates among undergraduate business students by analyzing the impact of
intellectual variance on business education. A quantile regression approach is utilized to
estimate whether the returns on certain student characteristics, most notably the variance in
intellectual ability as signaled by ACT score distribution across student cohorts in undergraduate
business programs, differ along the conditional distribution of their Major Field Test in Business
(MFT-B) test scores. A systematic examination of the relationship between academic ability
(using the ACT as a proxy) and academic achievement (measured by the MFT-B) found no
significant effects of either hetero- or homogeneity in academic ability variance on academic
achievement for high ability students. There was also no support for contentions that high ability
students might be disadvantaged by the presence of low ability colleagues. Quite interestingly a
positive and significant effect was found for lower ability students from 20th to 50th percentile
of the MFT-B distribution. While intellectual or academic ability, as signaled by the ACT,
certainly appears relevant in terms of individual achievement, there is no indication that an
admissions policy which creates cohorts with heterogeneity of innate intellectual ability has any
significant impact on the academic achievement high ability individuals and may in fact benefit
lower ability individuals within that cohort. Limitations and further research opportunities are
discussed.
Keywords: Academic Achievement, Intellectual Capability, Quantile Regression
Copyright statement: Authors retain the copyright to the manuscripts published in AABRI
journals. Please see the AABRI Copyright Policy at http://www.aabri.com/copyright.html.
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 2
INTRODUCTION
Anyone who has made even the briefest visit to a college of business faculty lounge or
attended one of their faculty meetings will have almost certainly encountered strongly held
opinions regarding the impact of student intellectual ability variance on teaching and academic
achievement. Strong contentions regarding the impact of students perceived to have intellectual
capabilities at the tails of the distribution are often expressed. Advocates for higher admission
standards may contend that low capacity students inhibit or hold back the learning of their
intellectually better endowed, capable classmates. Is that true? We seek to answer this question
by utilizing a large, multi-year sample collected from graduates of an Association to Advance
Collegiate Schools of Business (AACSB) accredited college of business at a Carnegie I, land
grant institution. This study seeks to extend previous lines of research which have identified
academic achievement determinants among undergraduate business students (e.g. Bielinska-
Kwapisz, et al., 2012) by analyzing the impact of intellectual variance on business education.
This is enabled by the use of a quantile regression approach which reveals effects at different
points of the distribution. The degree of variance in intellectual ability is measured by the
variance in ACT scores in a particular student cohort.
Whatever the stock of decision factors utilized in a business student’s selection of a major
field of study, those preferences and inclinations will eventually interact with opportunity,
options, and admission standards to create a cohort of students studying business with an
emphasis on a particular discipline at a college of business. The unifying factor in the resultant
cohort is, of course, the common interest in pursuing a certain field of study. However,
admission standards’ vagaries and imprecisions will inevitably create some degree of variation in
intellectual capability across the cohort. We study the previously un-researched impact of that
intellectual variation on the outcomes and academic achievement eventually realized by the
individuals in a post-secondary student cohort. The uniqueness of the data set used lies in the fact
that all students were tested by the same entry exam (ACT) and the same exit exam (MFT-B) but
were organized into cohorts (majors) during their courses of study.
The Major Field Test in Business (MFT-B) is published by the Educational Testing
Service (ETS), and is an assessment instrument designed for use in schools offering
undergraduate business programs. The ETS describes the MFT-B as being constructed according
to specifications developed and reviewed by subject matter expert committees so as to go beyond
the mere measurement of factual knowledge and to evaluate students’ ability to analyze and
solve problems, understand relationships, and interpret material from their major field of study
(Educational Testing Service [ETS], 2010). The ETS reports that the MFT-B was administered
to 181,488 individuals at 685 different institutions between 2006 and 2010. Martell (2007)
reported that, in 2006, 46% of business schools used the MFT-B test in their assessment of
students’ learning. The test contains 120 multiple-choice items covering the components of a
common body of knowledge for undergraduate business education in the following proportions:
accounting (15%), management (15%), economics (13%), finance (13%), marketing (13%),
qualitative analysis (11%), information systems (10%), legal and social environment (10%), and
international considerations of modern business operations (12% overlapping with the rest). The
scores on the MFT-B range from 120 to 200 and ETS reports the mean score as 153.1 with a
standard deviation of 14.1 from 2006 to 2010 (ETS, 2011).
A number of studies have examined the effect of students’ characteristics on performance
on the MFT-B (e.g. Bielinska-Kwapisz et al., 2012a, Bielinska-Kwapisz 2012b; Zeis et al.,
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 3
2009; Bycio and Allen, 2007; Bean and Bernardi, 2002; Stoloff and Feeney, 2002; Mirchandani,
et al., 2001; Allen and Bycio, 1997). However, all previous researchers have relied on estimation
using the Ordinary Least Squares (OLS) regression which estimates the mean effect of students’
characteristics on their MFT-B scores. While estimating how “on average” students’
characteristics effect performance is an important contribution to the understanding of the
dispositional factors which predict or explain academic achievement, it is also interesting to see
the effects at different points of the test score distribution. Quantile regression has two
particularly attractive features as compared to OLS. First, it allows the examination of covariates
effects on MFT-B scores along the entire MFT-B score distribution, providing a much more
detailed data description and analysis. Second, while OLS is sensitive to the presence of outliers,
quantile regression is more robust (Cameron and Trivedi, 2005).
The previously cited studies consistently identified ACT scores, GPA, gender,
motivation, and business major as having a significant predictive relationship with MFT-B
scores (for a complete literature review see Bielinska-Kwapisz et al., 2012a). While controlling
for all these factors, we extend this line of research by including the variance in ACT scores to
the list of predictors. In order to illuminate intellectual variance’s impact on achievement, we are
informed by the theoretical framework and extensive examination of this issue in the K-12 arena.
REVIEW OF LITERATURE
A review of the literature does not reveal any specific studies of the impact of intellectual
variation on performance in undergraduate business programs, or in higher education generally
for that matter. However, ability grouping, the practice of dividing students for instruction
according to their learning capacity is a common practice in the K-12 arena and has been
extensively studied. But after more than a half-century of analysis, ability grouping’s educational
impact remains in dispute. Proponents of ability grouping argue that it is an effective and
appropriate response to intellectual variation among students, and that it allows teachers to
provide appropriate instructional approaches for their students (National Education Association
[NEA], 1990; Wilson and Schmits, 1978). On the other hand, opponents contend that ability
grouping has significant undesirable consequence. They argue that when divided on the basis of
academic ability, classes also tend to be segregated by social and economic characteristics
(Oakes, 1990; Rosenbaum, 1976), and for a variety of reasons the low ability groups receive
instruction inferior to that provided to high ability students (Oakes, 1985; Page, 1991). As an
example, in a study of over 90 honors, regular, and remedial eighth and ninth grade English
classes, Gamoran et al. (1995) concluded that both the quality of instruction and student
outcomes were inferior in the low ability classes as compared to the high ability classes.
In 1997, Chicago public schools ended remedial classes and mandated college-
preparatory coursework for all students, without regard to student ability or inclination. Nomi
(2010), using an interrupted time series cohort design, concluded that although the policy
resulted in more students completing the ninth grade with algebra and English course credits,
failure rates increased, grades declined slightly, test scores did not improve, and students were no
more likely to enter college. On the other hand, in a within-class grouping meta-analysis of 20
effect sizes, Lou et al. (1996) found a significantly positive learning effect of homogeneous
ability grouping. Betts and Shkolnik (2000) found that a school policy of grouping students by
ability had little effect on average math achievement growth and little or no differential effects of
grouping for high-achieving, average or low-achieving students. Using a quasi-experimental
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 4
cohort design Burris et al. (2006) examined the effects of providing an accelerated mathematics
curriculum in heterogeneously grouped middle school classes and concluded that the probability
of completion increased significantly across all achievement categories. Finally, two major sets
of meta-analyses have been completed by Kulik and Kulik (1991) and Slavin (1987, 1990). By
performing a re-analysis of findings from all the studies included in these two sets of studies,
Kulik (2004) concluded that higher aptitude students usually benefit from ability grouping. The
effect is small when little or no adjustment to the curriculum is made, but larger if special classes
with accelerated curriculum are offered, while grouping has less influence on middle and lower
aptitude learners.
Clearly, academic outcomes determinates are multifactorial and complex, but the K-12
experience does not seem to provide us with a clear indication of the contribution of ability
grouping or intellectual variation on those outcomes. This unresolved debate, with logical
positions and evidence both pro and con, does identify an element of the post-secondary
educational setting which is presently unexamined and which, despite the absence of evidence,
elicits strong opinions, contentions, and advocacy and seems deserving of more systematic
consideration. The K-12 studies also provide examples of methodological approaches for that
examination. For instance, another set of literature compares the impact of student diversity in K-
12 education using international data. Michaelowa and Bourdon (2006) used the International
PISA data and found no support for a negative impact of intellectual heterogeneity on
performance, and in some cases the effect was even positive. Vandenberghe (2002) studied the
peer effects and concluded that for a given level of the average peer effect, ability heterogeneity
decreased students’ achievement in both math and science. These results are in almost perfect
opposite to those of Zimmer and Toma (2000). But despite what these studies lack in terms of
agreement on findings, they do have a common methodology in that they examined the effect of
diversity on an average student. The methodology is this study is similar to the one presented in
these international studies; however, through the use of quantile regression approach, it was
possible to determine the effects of heterogeneity on the high and low achieving students.
HYPOTHESES
The K-12 research on ability grouping, while still in process regarding the impact on academic
achievement, informs this study of the impact of intellectual heterogeneity in post-secondary
business education. Two hypotheses, one in regard to overall cohort impact and the other
examining the specific impact on high ability students are derived.
Hypothesis 1: Academic cohorts with heterogeneous distributions of academic ability will
have lower average levels of academic achievement than similarly situated academic
cohorts with homogeneous distributions of academic ability.
Hypothesis 2: Students with high academic ability in academic cohorts with
heterogeneous distributions of academic ability will perform less well than high ability
students in academic cohorts with homogeneous distributions of academic ability.
METHODS AND DATA
Quantile regression models the relation between a set of independent variables and specific
percentiles (or quantiles) of the dependent variable. Quantile regression was developed by
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 5
Koenker and Bassett (1978) and Koenker and Hallock (2001), and is based on the minimization
of weighted absolute deviations for estimating conditional quantile functions. Unlike linear
regression, in which the regression coefficient represents the change in the dependent variable
produced by a one unit change in the independent variable associated with that coefficient,
quantile regression parameters estimate the change in a specified quantile of the dependent
variable produced by a one unit change in the independent variable. Linear regression estimates
the mean value of the response variable for given levels of the predictor variables, while quantile
regression estimates the return across the conditional distribution. Quantile regression makes use
of the entire sample and is not equivalent to utilizing the dependent variable series of sub-
samples and applying OLS to sub-samples. Therefore, quantile regression is not the same as
dividing the data into different percentiles and then applying OLS to each percentile (e.g.
Hallock et al., 2008). The linear regression model in this study is in the following standard
form: 𝑦𝑖 = �̅�𝑖 �̅� + 𝜖𝑖 where 𝑦𝑖 is the dependent variable �̅�𝑖 is a vector of explanatory variables, �̅�
is a vector of their estimated coefficients, and 𝜖𝑖 is the independent and identically distributed
error term (Reichstein et al., 2010). The OLS estimator is found by minimizing the sum of the
squared residuals: min�̅�∈𝑅𝑘 [∑ (𝑦𝑖 − �̅�𝑖�̅�)2𝑛
𝑖 = 1 ]. On the other hand, the quantile regression
estimator is the vector β that minimizes:
min�̅�∈𝑅𝑘 [∑ 𝜏|𝑦𝑖 − �̅�𝑖 �̅�| + ∑ (1 − 𝜏)|𝑦𝑖 − �̅�𝑖�̅�|𝑖∈{𝑖:𝑦𝑖<�̅�𝑖 �̅�}𝑖∈{𝑖:𝑦𝑖≥�̅�𝑖 �̅�} ] where τ is the quantile
defined as 𝑄𝑌|𝑋(𝜏|𝑥) = 𝑖𝑛𝑓{𝑦: 𝐹𝑋|𝑌(𝑦|𝑥) ≥ 𝜏} in which τ is bounded between zero and one, and
y is a random sample from a random variable Y, which have the distribution function F (𝐹(𝑦) =𝑃(𝑌 ≤ 𝑦)). Notice that for τ = 0.5 the above equation becomes the absolute loss function
determining the median regression.
Quantile regression has been frequently applied to issues in labor economics when
examining wage differentials and wage discrimination (Fitzenberg et al., 2001; Garcia et al.,
2001; Buchinsky, 2001). Quantile regression was also employed to study K-12 education by
Eide and Showalter (1998) to estimate the relationship between school quality and student
performance. Levin (2001) estimated an educational production function to investigate the effect
of class size and peer effects. Prieto-Rodriguez et al., (2008) studied educational approaches in
14 European countries. In the only study that uses college data, Escudero et al. (2009) examined
the effects of gender, age, parent’s education, and private or public school on the number of
courses passed by students for accountancy and law students in Argentina.
In the context of this study, following Bielinska-Kwapisz et al. (2012a), ACT scores were
used as a proxy for general cognitive capability (Koenig et al., 2008), GPA as a measure of time
input and effort, along with students’ major field of study (accounting, finance, marketing, or
management). The specific learning function used in this study is a fixed effects model presented
as follows:
MFT-Bikt= β0 + β1 ACTikt + β2 GPAikt + β3 Maleikt + β4 ExCreditikt +ηk
+ δt
+ εikt (1)
where MFT-Bikt is the MFT-B score of student i in major k in term t; ACTikt is his/her ACT
score; GPAikt is his/her overall GPA; Maleikt is a binary variable that takes the value of one if a
student is male and zero if female; ExCreditikt is a binary variable that takes the value of one if a
student could receive extra points for a good performance; ηk are majors’ fixed effects; δt
are
cohort fixed effects; βi (i=1…4) are coefficients to be determined; and εikt is the error term.
Similar models were estimated in the previous literature on MFT-B scores by Bielinska-
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 6
Kwapisz, et al. (2012a); Zeis, et al. (2009); Rook and Tanyel (2009); Bycio and Allen (2007);
Bagamery et al. (2005); Black and Duhon (2003); Bean and Bernardi (2002); Mirchandani et al.
(2001); and Allen and Bycio (1997).
The unique contribution of this study is the examination of the impact of the effect of
heterogeneity in the ACT scores on the MFT-B scores; the issue none of the previous literature
explored. This study focuses on the composition of the students’ ACT scores rather than the pure
average of the ACT scores. To achieve this goal, a variance of students’ ACT scores was
included in Equation 1 in the following way:
MFT-Bikt= β0+ β1 ACTikt + β2 GPAikt + β3 Maleikt+ β4 ExCreditikt + β5VarACTkt + ηk
+ δt
+ εikt.(2)
Variances in the ACT scores in a given class and major were utilized (Michaelowa and Bourdon,
2006; Vandenberghe, 2002; Zimmer and Toma, 2000). The Levene test for the equality of
variances showed significant differences in variances between different majors-classes. The
variance ranges from 5.44 for accounting in 2007 to 16.23 for management in 2006.
Traditionally, Equation 1 was estimated by OLS. In this study, quantile regression, as described
above, was utilized to uncover previously unobserved heterogeneity in the returns to education
across quantiles.
The setting for the current study is an undergraduate college of business at a Carnegie
Research I, land grant university, which has held continuous AACSB accreditation for over 25
years. The college is predominantly Caucasian, with a very small population of international and
ethnic students. As part of an assessment of learning process, the college has administered the
MFT-B every semester to every graduating senior from the summer semester 2005 to spring
2011. Background data identified in the research were directly obtained from student records.
The total number of students in that population was 885 students. Full data, most notably MFT-B
and ACT scores, were available for 845 students, primarily attributable to the fact that transfer
students were not required to submit ACT scores. In addition, for each of the 845 students, the
data includes university grade point average measured at graduation (GPA), gender, major area
of study (finance, accounting, management, marketing), and graduating class (class 06, class 07,
class 08, class 09, class 11). Starting in the spring semester 2008, students received extra credit
points to incent their best efforts on the MFT-B (5 points for a 50th percentile score, 7.5 points
for 75th percentile, and so on). Table 1 (Appendix) reports the full list of variables that were
used, their definitions and descriptive statistics. Table 2 (Appendix) lists the observed
correlations. There are no very high (above ±0.5) correlations between any pair of independent
variables. Therefore, multicollinearity was of no consequence in the analyses. As the literature
from the K-12 experience suggests, there may be a different effect of cohort heterogeneity on
high and low achieving students. For example, high ACT variance may influence MFT-B scores
differently for high-performing students than for low-performing students. However, since the
previous literature produced mixed evidence, the sign and significance of this variable has to be
determined empirically.
RESULTS
If, on average, increased cohort heterogeneity leads to reduced MFT-B scores, it would
be expected that cohorts with a high ACT dispersion will return lower MFT-B scores. Figure 1
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 7
(Appendix) relates average MFT-B cohort scores to the heterogeneity of their ACT scores and
provides no supportive for hypothesis 1.
However, the relationship has to be further investigated by controlling for other input
variables through the regression analysis. Table 3 (Appendix) reports estimated coefficients for
the model described in Equation 2 estimated by the ordinary least squares (OLS) and, therefore,
reports estimate at the mean.
The model explains about 49% of the variation in the MFT-B scores. The results from the
estimated regression show no significant effect of ACT dispersion on the average MFT-B scores.
Thus no support was found for hypothesis 1, that higher cohort heterogeneity of academic ability
as measured by ACT scores retards (or benefits) the average MFT-B scores.
Other coefficients are consistent in significance and magnitude with those reported in
previous literature, for example in Bielinska-Kwapisz et al. (2012a) and Black and Duhon
(2003). In particular, higher ACT score and GPA positively influence MFT-B scores. Male
students score, on average, four points higher on the MFT-B (Bielinska-Kwapisz and Brown,in
press). Finally, management and marketing students score, on average, lower than accounting
students (accounting and cohort 2011 dummies were dropped from the equation to avoid perfect
multicollinearity).
The shortcoming of the OLS regression is that it estimates coefficients at the mean values
but the effect of the dispersion may influence top or bottom students differently than the average.
As commonly suggested, high ACT students may be at an academic disadvantage in class with
low performing students, or low ACT students may benefit by being in class with higher
potential students. However, OLS cannot be used to answer these hypotheses; therefore quantile
regression is utilized. In Table 4 (Appendix) , estimated coefficients at 0.05, 0.25, 0.50, 0.75,
0.95 quantiles of the MFT-B scores distribution are reported.
Of particular interest in this study are the coefficients of the dispersion. The coefficients
for the top 5% (95th
) and top 25% (75th
) MFT-B students are small and insignificant. Clearly, the
top students are not disadvantaged by being in heterogeneous cohorts. Thus there is no support
for hypothesis 2 that students with high academic ability in academic cohorts with heterogeneous
distributions of academic ability will perform less well than high ability students in academic
cohorts with homogeneous distributions of academic ability.
Turning to lower performing students, there is also no significant observed effect of
higher dispersion on the very bottom 5% of the students. However, quite interesting the
coefficient is positive and significant from the bottom 20% students (20th
) to the median of the
distribution (50th
). In fact, the impact is the largest at 25 percentile and decreases afterward.
Clearly, these relatively low academic ability students benefited from being in heterogeneous
cohorts.
Turning to the effect of other variables, the quantile regression results suggest some
important differences across diverse points in the conditional distribution of MFT-B scores. The
intercept is increasing along the MFT-B distribution, suggesting that the unexplained portion of
the variation in MFT-B scores are the highest at the top of the distribution.
In this study ACT scores are used to represent cognitive intelligence and academic
ability. A 1% increase in an ACT score is associated with an MFT-B score that is 23% higher at
the median (50th
), but only 19% higher at the top (95th
) of the MFT-B distribution (all elasticities
evaluated at sample means).Therefore, academic ability seems to have it biggest impact at the
middle of the MFT-B distribution.
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 8
The larger the GPA coefficient, the greater is the impact of business education on
students. If coefficients are not stable through the quantiles (i.e. changing along the MFT-B
distribution), then there is inequality in education along the distribution. GPA should reflect
students’ general business knowledge and how well the academic business program prepares
students for the exam. GPA slightly declines along the MFT-B distribution (Table 4). A 1%
increase in GPA is associated with an MFT-B score that is 15% higher at the bottom of the
distribution (5th
) and only slightly declines to 13% at the top of the distribution (95th
). Therefore,
the GPA impact is pretty stable across the distribution.
In the study samples, students who took the MFT-B in spring 2008 and subsequently,
received extra credit points to incent their best efforts on the MFT-B. The OLS results suggest a
significant and positive effect of the extra credit. On average, the scores increased by 2.5 points
(Table 3). The results from the quantile regression (Table 4) suggest that effect of extra credit
was not the same at the considered points of distribution. The coefficients were significant at the
top of the distribution only (75th
and 95th
percentile). The effect was the largest at the 5% top of
the MFT-B distribution: 3.2 points (Bielinska-Kwapisz and Brown (in press) analyzed the effect
of extra credit in more detail).
On average, male students in the study sample outperformed female students by 4.57
points on the MFT-B exam (Table 3), a phenomenon reported in many previous studies.
However, the effect is much smaller at the bottom of the distribution (2.64) and the first quartile
(3.99) (Table 4). The effect is the largest at the third quartile (5.74) and at the median (4.82)
(Bielinska-Kwapisz and Brown (2in press) analyzed the effect of gender in more detail).
Interestingly, the lower score for management students compared to accounting was not
significant at the top 5% of the distribution. Finance students significantly outperformed
accounting students at the top 5% of the distribution by 4.06 points (even though the OLS
coefficient is not significant).
CONCLUSION
This empirical study examines the impact of the degree of variation in academic ability
on the academic achievement of student cohorts and the highest ability students in those cohorts.
Significant differences in the variance of academic ability (utilizing variation in ACT scores as a
proxy) were observed across field of study cohorts. Quantile regression was utilized to examine
the impact of this dispersion, while controlling for other dispositional factors, on academic
outcomes as measured by scores on the MFT-B. A systematic examination of that relationship
found no significant effects of either hetero- or homogeneity in academic ability variance of
academic cohorts on an average student. There was also no support for contentions that high
ability students might be disadvantaged by the presence of low ability colleagues. An unexpected
positive and significant academic achievement effect was found: from 20th
to 50th
percentile of
the MFT-B distribution for lower ability students in heterogeneous cohorts.
The findings were determined through the use of quantile regression which examined
relationships at various locations on the distribution curve. Previous research on the determinants
of MFT-B scores estimated coefficients at the mean. However, it is useful to see the
determinants of the MFT-B scores examined at different points of the distribution. The average
impact may significantly differ from the impact at the top or a bottom of the distribution.
Quantile regression would seem to substantially improve assessment of learning accuracy and
contribute to a better and much more precise understanding of the dispositional factor
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 9
contributions which impact academic achievement as compared to estimates of the coefficients
on sample means obtained by the use of OLS.
The results of this empirical study do not provide any support for the oft-stated
contention that intellectually less well-endowed students’ presence in undergraduate business
classrooms might inhibit the academic achievement of high capacity students. Probably the main
implication of the study results can be applied to the methods and procedures by which students
are assigned or admitted into universities, colleges, or particular fields of study.
While intellectual or academic ability, as signaled by the ACT, certainly appears relevant
in terms of individual achievement, there is no indication that an admissions policy which creates
cohorts with heterogeneity of innate intellectual ability has any significant impact on the
academic achievement of either the cohort of the cohort’s high ability students. In fact, study
results suggest that a heterogeneous academic ability cohort may in fact benefit the lower ability
individuals in that cohort without any deleterious impact on other members. Those looking for
explanations for frustrations or academic performance below expectations should look at factors
other than the presence of low ability students.
The findings of this study are of course bound by the validity of the proxy measures
utilized: ACT as a measure of intellectual ability; GPA as a measure of time input and effort; and
MFT-B scores as a measure of academic achievement. Generalization of the findings will be
enhanced by replications in other institutional and disciplinary settings, something which is
strongly hoped for and encouraged.
REFERENCES
Allen, J.S. and Bycio, P. (1997), “An evaluation of the Educational Testing Service Major Field
Achievement Test in Business”, Journal of Accounting Education, Vol. 15, pp. 503-514.
Bagamery, B.D., Lasik, J.J., and Nixon, D.R. (2005), “Determinants of success on the ETS
Business Major Field Exam for students in an undergraduate multisite regional university
business program”, Journal of Education for Business, Vol. 81 No. 1, pp. 55-63.
Bean, D.F. and Bernardi, R.A. (2002), “Performance on the Major Field Test in Business: The
explanatory power of SAT scores and gender”, Journal of Private Enterprise, Vol. 17/18,
pp. 172-178.
Betts, J.R. and Shkolnik, J.L. (2000), “The effects of ability grouping on student math
achievement and resource allocation in secondary schools”, Economics of Education
Review, Vol. 19, No. 1, pp. 1-15.
Bielinska-Kwapisz, A., Brown, F.W., and Semenik, R.J. (2012a), “Is higher better?
Determinants and comparisons of performance on the Major Field Test-Business”,
Journal of Education for Business. 87 (3), 1-11.
Bielinska-Kwapisz, A., Brown, F.W. and Semenik, R.J. (2012b), “Interpreting standardized
assessment test scores and setting performance goals in the context of student
characteristics: The case of the major field test in business”, Journal of Education for
Business. 87 (1), 7-13
Bielinska-Kwapisz, A. and Brown, F.W. (in press), “Differential Gender Performance on the
Major Field Test–Business”, Journal of Education for Business.
Black, H.T. and Duhon, D.L. (2003), “Evaluating and improving student achievement in
business programs: The effective use of standardized assessment tests”, Journal of
Education for Business, Vol. 79, pp. 90-98.
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 10
Buchinsky, M. (2001), “Quantile regression with sample selection: Estimating women's return to
education in the U.S.”, Empirical Economics, Vol. 26, No. 10, pp. 87-113.
Burris, C.C., Heubert, J., and Levin, H. (2006), “Accelerating mathematics achievement using
heterogeneous grouping”, American Educational Research Journal, Vol. 43, No. 1, pp.
103-134.
Bycio, P. and Allen J.S. (2007), “Factors associated with performance on the Educational
Testing Service (ETS) Major Field Achievement Test in Business (MFT-B)”, Journal of
Education for Business, Vol. 82, No. 4, pp. 196-201.
Cameron, C.A. and Trivedi, P.K. (2005), Microeconomics: Methods and application, Cambridge
University Press, London.
Educational Testing Service (ETS) (2011), “Individual students test score distribution”, available
at http://www.ets.org/s/mft/pdf/2010/business4cmf.pdf (accessed 8 February 2011).
Educational Testing Service (ETS) (2010), “About Scores and Reports”, available at
https://www.programworkshop.com/6.0.0.0/
home.aspx?skin=MMIandsc=I004E0F604E403A801D3F38602BB20A3D75A17670
(accessed 7 January 2010).
Eide, E. and Showalter, M.H. (1998), “The effect of school quality on student performance: A
quantile regression approach”, Economics Letters, Vol. 58, No. 3, pp. 345-350.
Escudero, W.S., Giovagnoli, P.I., and Porto, A. (2009), “The effects of individual characteristics
on the distribution of college performance”, ECONÓMICA, Vol. LV, pp. 99-130.
Fitzenberg, B., Hujer, R., MaCurdy, T.E., and Schnabel, R. (2001), “Testing for uniform wage
trends in West-Germany: A cohort analysis using quantile regressions for censored data”,
Empirical Economics, Vol. 26, pp. 41-86.
Garcia, J., Hernández, P.J., and Nicolás, Á.L. (2001), “How wide is the gap? An investigation of
gender wage differences using quantile regression”, Empirical Economics, Vol. 26, pp.
149-167.
Gamoran, A., Nystrand, M., Berends, M., and LePore, P.C. (1995), “Pan organizational analysis
of the effects of ability grouping”, American Educational Research Journal, Vol. 32, pp.
687-716.
Hallock, K.F., Madalozzo, R., and Reck, C.G. (2008), “CEO pay-for-performance heterogeneity:
Examples using quantile regression”, The Financial Review, Vol. 45, pp. 1-19.
Koenig, K.A., Frey, M.C., and Detterman, D.K. (2008), “ACT and general cognitive ability”,
Intelligence, Vol. 36, pp. 153-160.
Koenker, R. and Bassett, Jr., G. (1978), “Regression quantile”, Econometrica, Vol. 46, No. 1, pp.
33-50.
Koenker, R. and Hallock, K.F. (2001), “Quantile regression”, Journal of Economic Perspectives,
Vol. 15, No. 4, pp. 143-156.
Kulik, J.A. (2004), “Meta-analytic studies of acceleration”, in Colangelo, N., Assouline, S. and
Gross, M.U.M. (Eds.), A nation deceived: How schools hold back America’s brightest
students, The University of Iowa, The Connie Belin and Jacqueline N. Blank
International Center for Gifted Education and Talent Development, Iowa City, IA, Vol.
2, pp. 13-22.
Kulik, J.A., and Kulik, C.L.C. (1991), “Ability grouping and gifted students,” in Colangelo, N.
and Davis, G.A. (Eds.), Handbook of Gifted Education, Allyn and Bacon, Boston, MA,
pp. 178-196.
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 11
Levin, J. (2001), “For whom the reductions count: A quantile regression analysis of class size
and peer effects on scholastic achievement”, Empirical Economics, Vol. 26, No. 1, pp.
221-246.
Lou, Y., Abrami, P.C., Spence, J.C., Poulsen, C., Chambers, B., and d’Apollonia, S. (1996),
“Within-class grouping: A meta-analysis”, Review of Educational Research, Vol. 66, No.
4, pp. 423-58.
Martell, K. (2007). “Assessing student learning: Are business schools making the grade?”,
Journal of Education for Business, Vol. 82, No. 4, pp. 189-195.
Michaelowa, K. and Bourdon, J. (2006), “The impact of student diversity in secondary schools:
An analysis of the International PISA data and implications for the German education
system”, research paper, Hamburg Institute of International Economics (HWWI),
Hamburg, Germany, pp. 3-2.
Mirchandani, D., Lynch, R., and Hamilton, D. (2001), “Using the ETS Major Field Test in
Business: Implications for assessment”, Journal of Education for Business, Vol. 77, No.
1, pp. 51-55.
National Education Association (NEA) (1990), “Academic tracking”, report of the NEA
Executive Committee Subcommittee on Academic Tracking, Washington, D.C.
Nomi, T. (2010), “The unintended consequences of an algebra-for-all policy on high-skill
students: The effects on instructional organization and students’ academic outcomes”,
paper presented at the Fall Research Conference of the Association for Public Policy
Analysis and Management, Boston, MA.
Oakes, J. (1985), Keeping track: How schools structure inequality. Yale University Press, New
Haven, CT.
Oakes, J. (1990), Multiplying inequalities: The effects of race, social class, and tracking on
opportunities to learn mathematics and science, RAND, Santa Monica, CA.
Page, R.N. (1991), Lower track classrooms: A curricular and cultural perspective, Teachers
College Press, New York, NY.
Prieto-Rodriguez, J., Barros, C., and Vieira, J. (2008), “What a quantile approach can tell us
about returns to education in Europe”, Education Economics, Vol. 16, No. 4, pp. 391-
410.
Reichstein, T., Dahl, M.S., Ebersberger, B., and Jensen, M. (2010), “The devil dwells in the tails:
A quantile regression approach to firm growth”, Journal of Evolutionary Economics, Vol.
20, No. 2, pp. 219-231.
Rook, S.P. and Tanyel, F.I. (2009), “Value-added assessment using the Major Field Test in
Business”, Academy of Educational Leadership Journal, Vol. 3.
Rosenbaum, J.E. (1976), Making inequality: The hidden curriculum of high school tracking,
Wiley, New York, NY.
Slavin, R.E. (1987), “Ability grouping and student achievement in elementary schools: A best-
evidence synthesis”, Review of Educational Research, Vol. 57, pp. 293-336.
Slavin, R.E. (1990), “Achievement effects of ability grouping in secondary schools: A best-
evidence synthesis”, Review of Educational Research, Vol. 60, pp. 471-499.
Stoloff, M., and Feeney, K.J. (2002), “The Major Field Test as an assessment tool for an
undergraduate psychology program,” Teaching of Psychology, Vol. 29, No. 2, pp. 92-98.
Vandenberghe, V. (2002), “Evaluating the magnitude and the stakes of peer effects analysing
science and math achievement across OECD”, Applied Economics, Vol. 34, No. 10, pp.
1283-1290.
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 12
Wilson, B.J. and Schmits, D.W. (1978), “What's new in ability grouping?”, Phi Delta Kappan,
Vol. 59, pp. 535-536.
Zeis, C., Waronska, A., and Fuller, R. (2009), “Value-added program assessment using
nationally standardized tests: Insights into internal validity issues”, Journal of Academy
of Business and Economics, Vol. 9, No. 1, pp. 114-128.
Zimmer, R.W. and Toma, E.F. (2000), “Peer effects in private and public schools across
countries”, Journal of Policy Analysis and Management, Vol. 19, No. 1, pp. 75-92.
APPENDIX
TABLE 1
Definitions and Descriptive Statistics (n=845)
Variable Description Mean Std
Dev.
Minimum Maximum
MFT-B Student MFT-B score on a scale of 120 to
200
160.86 12.16 129 194
ACT Student ACT score on a scale of 1 to 36 23.39 3.42 14 34
GPA GPA measured at the time of graduation 3.14 0.38 2.26 4.00
Var(ACT) Variance of ACT scores in a given major
and class
11.15 2.44 5.44 16.23
Male Binary variable = 1 if male 0.55 0 1
ExCredit Binary variable = 1 if extra credit was
offered
0.43 0 1
Finance Binary variable = 1 if finance major 0.17 0 1
ACCT Binary variable = 1 if accounting major 0.21 0 1
MGMT Binary variable = 1 if management major 0.34 0 1
MKTG Binary variable = 1 if marketing major 0.28 0 1
Class
2006
Binary variable = 1 if class 2006 0.19 0 1
Class
2007
Binary variable = 1 if class 2007 0.18 0 1
Class
2008
Binary variable = 1 if class 2008 0.22 0 1
Class
2009
Binary variable = 1 if class 2009 0.22 0 1
Class
2011
Binary variable = 1 if class 2011 0.19 0 1
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 13
TABLE 2
Correlations (n=845)
Variable 1 2 3 4 5 6 7 8 9 10
1 MFT 1.00
2 ACT 0.57 1.00
3 VarACT 0.02 -0.04 1.00
4 GPA 0.45 0.46 -0.02 1.00
5 Male 0.14 -0.04 0.12 -0.20 1.00
6 ExCredit 0.08 -0.03 -0.23 -0.01 0.07 1.00
7 ACCT 0.21 0.17 -0.24 0.21 -0.16 0.06 1.00
8 FIN 0.26 0.08 0.01 0.03 0.15 0.03 -0.23 1.00
9 MGMT -0.15 -0.12 0.45 -0.09 0.13 -0.06 -0.37 -0.32 1.00
10 MKTG -0.25 -0.09 -0.27 -0.12 -0.12 -0.01 -0.32 -0.28 -0.45 1.00
TABLE 3
Ordinary Least Squares (OLS) Estimates (n=845)
Estimate t stat
Intercept 96.25 28.26
ACT 1.48 14.79
VarACT 0.24 1.52
GPA 7.84 8.56
Male 4.57 7.17
ExCredit 2.50 2.53
FIN 1.81 1.77
MGMT -5.17 -5.37
MKTG -5.91 -6.63
Class06 3.43 3.04
Class07 3.39 3.21
Class08 1.44 1.43
Class09 2.71 2.37
R-squared: 0.494
Adjusted R-squared: 0.487
Research in Higher Education Journal
Intellectual heterogeneity’s impact, page 14
TABLE 4
Quantile Regression (n=845)
Variable 5th
25th
50th
75th
95th
Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat Coeff. t-stat
Intercept 87.17 19.03 86.98 20.67 90.78 25.04 107.77 20.54 122.42 25.42
ACT 1.40 13.66 1.46 12.33 1.60 14.55 1.42 9.69 1.32 10.84
VarACT 0.10 0.78 0.48 2.32 0.31 1.86 0.31 1.18 -0.27 -0.96
GPA 7.93 9.36 8.02 7.21 7.92 7.96 6.93 5.12 6.82 6.55
Male 2.64 4.83 3.99 5.34 4.82 6.81 5.74 6.09 4.06 4.87
ExCredit 1.77 0.80 1.18 1.10 1.94 1.50 3.03 1.84 3.20 2.01
FIN -0.81 -0.58 2.71 1.59 2.64 2.16 0.30 0.20 4.06 3.83
MGMT -3.96 -3.91 -4.66 -3.79 -4.33 -3.97 -8.79 -5.53 -2.15 -1.35
MKTG -6.06 -6.64 -4.86 -5.49 -5.06 -5.73 -8.54 -5.76 -4.95 -3.36
Class06 2.00 0.91 4.40 3.69 4.69 4.05 2.58 1.62 2.27 1.93
Class07 3.15 1.34 3.19 2.30 3.36 3.13 2.37 1.26 3.83 3.69
Class08 0.61 0.41 2.40 2.21 2.59 2.07 -0.05 -0.03 1.75 1.12
Class09 3.17 2.76 3.94 2.75 3.88 2.43 1.89 0.92 2.49 1.09
FIGURE 1. Average MFT-B Class Score and Variance in ACT Scores.
152
154
156
158
160
162
164
166
168
170
5 7 9 11 13 15 17
Me
an M
FT-B
sco
re
Variance in ACT scores
top related